LIBRO DE ACTAS (pdf) - Universidad de Sevilla

More documents

Recommendations

Info

Melodic Transcription of Flamenco Singing from Monophonic and Polyphonic Music Recordings Melodic Transcription of Flamenco Singing from Monophonic and Polyphonic Music Recordings EMILIA GÓMEZ JORDI BONADA JUSTIN SALAMON Music Technology Group, Universitat Pompeu Fabra, Barcelona, Spain {emilia.gomez, jordi.bonada, justin.salamon}@upf.edu - http://mtg.upf.edu Abstract We propose a method for the automatic transcription of flamenco singing from monophonic and polyphonic music recordings. Our transcription system is based on estimating the fundamental frequency (f0) of the singing voice, and follows an iterative strategy for note segmentation and labelling. The generated transcriptions are used in the context of melodic similarity, style classification and pattern detection. In our study, we discuss the difficulties found in transcribing flamenco singing and in evaluating the obtained transcriptions, we analyze the influence of the different steps of the algorithm, and we state the main limitations of our approach and discuss the challenges for future studies. 1. Introduction 1.1 Motivation Flamenco is a music tradition originating mostly from Andalusia in southern Spain. The origin and evolution of the different flamenco styles (palos) and variants have been studied by different disciplines, including ethnomusicology, literature and anthropology. A frequent topic of discussion among flamenco scholars concerns the comparison of different performances and the precise definition of styles and variants. Current Music Information Retrieval (MIR) technologies could provide a different perspective to this debate. First, they might help to set up a standard methodology for flamenco description and comparative analysis, and support the formalization of expert knowledge. Second, they might facilitate the study of large music collections. Flamenco music germinated and nourished mainly from the singing tradition. Accordingly, the singer’s role soon became dominant and fundamental. Often, the singer is accompanied by the flamenco guitar; other flamenco instruments include claps, rhythmic feet and percussion. This work focuses on melodic description, and is driven by the research hypothesis that each flamenco style is characterized by a certain melodic skeleton or contour, which can be subject to ornamentation and variation (Donnier, 1997; Mora et al., 2010). The aim of this work is therefore to provide a method to extract detailed melodic transcriptions from audio signals, which can then be processed for ornament detection and further simplified to obtain the overall melodic contour. We will deal with a cappella singing (martinete and debla styles) and singing with guitar accompaniment (fandango style). 1.2 Flamenco and its musical transcription Because of its oral transmission, there are no written scores in flamenco music. Flamenco experts have put much effort into generating manual transcriptions after listening to live performances or field recordings, as a means to catalogue, classify and imitate the most relevant performers and their stylistic traits (Hurtado and Hurtado, 1998); (Hurtado and Hurtado, 2002); (Fernández, 2004); (Hoces, 2011). As pointed out by Toiviainen & Eerola (2006) and Lesaffre et al. (2004) in other contexts, manual analyses provide very accurate and expert information, but they are very time consuming and might be subjective or prone to errors. This is also the case in flamenco, due to two main reasons. First, there is a disagreement on the most adequate transcription methodology; For instance, Donnier (1997) proposed the adaptation of plainchant neumes. Hurtado and Hurtado (1998, 2002), on the contrary, forcefully argue for the use of Western notation. Second, even if we agree on the use of a certain format, there is a degree of subjectivity in the transcription process, given the high degree of ornamentation in flamenco music. 199
Melodic Transcription of Flamenco Singing from Monophonic and Polyphonic Music Recordings 1.3 Automatic transcription of sung melodies Automatic transcription is one of the main research challenges in the field of sound and music computing. It consists in computing a symbolic musical representation (in terms of Western notation) from a given musical performance (Klapuri, 2006). For monophonic music material, the obtained transcription relates to the melody (Gómez et al., 2003) and in polyphonic music material there is an interest in transcribing the predominant melodic line (Klapuri, 2006). Transcription systems can provide melodic descriptors at different levels. The main melodyrelated Low-level features are energy, associated with loudness, and fundamental frequency (f0) related to its perceptual correlate, pitch. From now on, we will use the term pitch to refer to f0. In a higher structural level, audio streams are segmented into notes, and their duration and pitch provide a symbolic representation. This representation can be the input to higher-level music analyses, e.g. ornament detection, melodic contour extraction or key or scale analysis. Current systems for automatic transcription are usually composed of three different stages: low-level (frame-based) descriptor extraction, note segmentation and note labelling. When dealing with monophonic music signals, existing transcription systems provide satisfying results for a great number of musical instruments. Although we find some successful approaches for singing voice (Mulder et al. 2003; Ryynänen, 2006), it is still one of the most complex instruments to transcribe, even in a monophonic context. This is due to several factors, such as the continuous character of the human voice and the variety of pitch ranges and timbre. This results in difficulties in obtaining correct f0 estimations, detecting note transitions and labelling notes in terms of pitch or duration. When dealing with polyphonic music signals, current state-of-the-art algorithms for predominant f0 estimation yield an overall accuracy around 75% according to the 2011 edition of the Music Information Retrieval Evaluation eXchange (MIREX). Moreover, audio onset detection methods yield an average F-measure around 0.78 (MIREX). This F-measure is obtained for a mixed dataset of 85 files, but if we just consider the 5 tested singing voice excerpts, the maximum F-measure is 0.47. In addition, current approaches are oriented towards mainstream popular music. This leads us to the question of how would these algorithms perform for, e.g. traditional music, and more particularly, flamenco singing. Additional challenges in flamenco transcription arise from the quality of existing recordings, the acoustic and expressive particularities of singing, its ornamental and improvisational character and the yet to be formalized musical structures employed (Mora et al., 2010). 2. Selected approach Figure 1 shows an overall diagram of the proposed system, which is based on the one described in (Janer et al., 2008). It consists of four main steps: low-level feature extraction (fundamental frequency, energy and spectral features), tuning frequency estimation, transcription into short notes, and an iterative process involving note consolidation and refinement of the tuning frequency. 200
Page 2:
Las fronteras entre los géneros: f
Page 5 and 6:
Reservados todos los derechos. Ni l
Page 7 and 8:
ÍNDICE Preface …. …………
Page 10:
Preface The III Interdisciplinary C
Page 13 and 14:
Ciencias de la Salud. De estos caso
Page 15 and 16:
Hacia una revisión del concepto
Page 17 and 18:
Page 19 and 20:
Page 21 and 22:
Page 23 and 24:
Page 25 and 26:
Page 28 and 29:
El Romance de La loba parda y la Bu
Page 30 and 31:
Page 32 and 33:
Page 34 and 35:
Page 36 and 37:
Page 38 and 39:
Page 40 and 41:
Flamenco - Jazz: una perspectiva an
Page 42 and 43:
Page 44 and 45:
Page 46 and 47:
Page 48 and 49:
Page 50 and 51:
Page 52 and 53:
EL canto (cante) al Cristo de la C
Page 54 and 55:
Page 56 and 57:
Page 58 and 59:
Page 60:
Page 63 and 64:
Estudio de métodos para transcribi
Page 65 and 66:
Page 67 and 68:
Page 69 and 70:
Page 71 and 72:
Page 73 and 74:
El zapateado flamenco: naturaleza f
Page 75 and 76:
Page 77 and 78:
Page 79 and 80:
Page 81 and 82:
PERFIL FLAMENCO: ESTEBAN SANLÚCAR
Page 83 and 84:
Page 86 and 87:
Federico García Lorca y la música
Page 88 and 89:
Page 90 and 91:
Page 92 and 93:
Page 94 and 95:
Page 96:
Page 99 and 100:
El flamenco como expresión de lo s
Page 101 and 102:
Page 103 and 104:
Page 105 and 106:
Page 107 and 108:
Nuevas propuestas para la historia
Page 109 and 110:
Page 111 and 112:
Page 113 and 114:
Page 115 and 116:
The legacy of Félix Fernández Gar
Page 117 and 118:
The legacy of Félix Fernández Gar
Page 119 and 120:
118
Page 121 and 122:
¿Tradición o renovación?: Antoni
Page 123 and 124:
Page 125 and 126:
Page 127 and 128:
Page 129 and 130:
Page 131 and 132:
Page 133 and 134:
Page 135 and 136:
Page 137 and 138:
Page 139 and 140:
138
Page 141 and 142:
El flamenco como epistemología par
Page 143 and 144:
Page 145 and 146:
Page 147 and 148:
Page 149 and 150: El flamenco como epistemología par
Page 161 and 162: La audición en los más pequeños
Page 169 and 170: 168
Page 171 and 172: Estrategias didácticas para el uso
Page 181 and 182: Arturo Pavón, recorrido diacrónic
Page 193 and 194: A Robust Audio Fingerprinter Based
Page 199: A Robust Audio Fingerprinter Based
Page 203 and 204: Melodic Transcription of Flamenco S
Page 213 and 214: Retentional Syntagmatic Network, an
Page 219 and 220: 218
Page 221 and 222: Antipattern discovery in Basque fol
Page 227 and 228: Automatic Detection of Melodic Patt
Page 235 and 236: 234
Page 237 and 238: Does singing style correlate to soc
Page 239: Does singing style correlate to soc
show all

LIBRO DE ACTAS (pdf) - Universidad de Sevilla

Create successful ePaper yourself

Delete template?

Save as template?